Re:Google said it had successfully ‘slowed down’ European pr...
by jboy55 on Monday October 25, 2021 @03:32PM
(#28992087)
My point is, that if you trained an AI on Parisian French, people in Quebec would be annoyed but could adapt to the differences if they wanted to use the system. People in southern Germany may be annoyed they have to speak high German, but they know it. There are very few people, if any, who only understand Cajun French and the issue.
Also, the biggest difference would be the pronunciation differences vs words being different. The word differences are easily modeled after any ASR, if the developer understands the differences and plans for it in their NLU models.
In the US the dialects are very similar and usually separated mostly by accent.
When dealing with different countries, entire groups of words do not mean the same thing because either (a) the language has changed, or (b) those things literally did not exist at the time the split occurred. Training an AI on Parisian French will result in an utterly broken experience for Haitian French and Cajun French.